Learning a Second Language with a Videogame1
Marilisa Amoia, Claire Gardent, Laura Perez-Beltrachini
INRIA Nancy Grand Est, CNRS, University of Nancy (France)
[email protected], [email protected], [email protected]
Abstract
In this paper, we present an experiment designed to evaluate I-Fleg, a serious game prototype for
second language acquisition we implemented in our laboratory. Specifically, our experiment focuses
on teaching and training French elementary vocabulary to German primary school children.
1. Introduction
Young children, when faced with learning a second language, seem to apply similar cognitive
mechanisms as those underlying first language acquisition by native speakers (cf. [5]). This
observation particularly apply to very young children such as pre-school and primary school children.
Therefore, at this stage the exposure to the second language assume a relevant role in second
language teaching and many educators [6] argue that full-immersion techniques are the most
appropriate for primary school children.
Accordingly, in second language teaching classes at primary school level natural methods, e.i.
teaching methodologies fostering interactions rather than providing instructions, are preferred to the
formal methods that are applied in later educational stages. Hence, classroom teaching is generally
based on rituals recreating real, concrete situations, that are used as a context for learning thereby
providing learning through meaning communication and cultural exposure.
Another important issue in second language teaching at primary school is the playful character of the
learning experience. Many authors (e.g. [9], [3] and [1]) point out that games are a central mean to
convey linguistic knowledge to children as they allow to learn and use language in context and
promote the communicative as well as the natural, spontaneous use of language. Further, as has
been often noted [4], games might help maintain a high attention level and also promote both interest
and engagement of learners.
In this paper, we present an experiment designed to evaluate the impact of I-Fleg, an interactive ITS
for second language acquisition, on teaching French as a second language to German primary school
children.
Second language teaching in German primary schools is not performance-oriented. Its aim is rather to
convey sensitivity to the phonemic structure of the second language to young children. Therefore, in
the traditional classroom listening and understanding activities (i.e. passive language learning) are
privileged with respect to activities targeting the acquisition of the second language grammar rules
such as spelling, orthography exercises or essay-writing activities.
In our experiment we studied how, a very difficult task for German primary school children such as
learning new words and their spelling in the second language, can be facilitated by the 3D interactive,
game-like environment provided by our tutorial prototype system.
1
The research presented in this paper was partially supported by the European Fund for Regional Development
within the framework of the INTERREG IV A Allegro Project.
The paper is organized as follows. In section 2, we presents I-Fleg, the game-like prototype for second
language acquisition we have implemented. Then in section 3, we describe the experiment we
designed to study the impact of I-Fleg on primary school children learning French. Finally in section 4,
we conclude discussing the results of the evaluation.
2. I-Fleg: A 3D Video Game for learning French
I-Fleg is a language learning prototype we have developed at our laboratory. It has been implemented
as a 3D serious game for teaching and training French as a second language. The core of I-Fleg is
implemented in Java, however for rendering the 3D graphics of the game virtual world we have
integrated it with the graphical engine provided by Second Life (see [2] for more details).
The game scenario is a virtual house in which the player’s avatar can freely navigate, explore the
rooms and interact with their objects, by means of chatting (i.e.. text input), touch events and change
of position, and by doing so they train their French. The system provides a tutorial agent, that is a
window displaying guidelines for the training activity at hand or feedback on learner answer.
In the experiment, the tutorial system was adapted to interact with young children. The game consists
of two phases: tutorial and a training phase. In the tutorial phase the player explores the house and
learns new vocabulary by touching the objects in the rooms. The system provides both written and
audio output describing the name of the object and its pronunciation. In the training phase, the system
asks the player the name of the touched object. In order to score credit points, the player has to type
the name of the object in the chat box. The goal of the game is to collect the highest possible number
of credit points. To boost the playful aspect of the game, we use 5 blinking stars, each star starts to
blink when the player has scored 3 credit points, so that the ultimate goal of the game is to get 5 stars
blinking. Figure 1 shows an example scenario from the game.
Figure 1. I-Fleg: An Example Scenario.
In order to foster children motivation and to meet the German primary school second language
teaching guidelines, where the acquisition of a language feeling and of a sensitivity to the phonemic
structure of the second language has a higher rank than correct word spelling, in our system we
implemented a feedback strategy that also accept partially correct answers, i.e. answers that are
misspelled but can still be understood.
We compared different word distance metrics and find that the Levenshtein distance best fits the kind
of data at hand. The French words learned at primary school level are in fact generally very short.
Thus, the system accepts an answer as correct, if its Levenshtein distance from the expected answer
generated by the system is greater than a threshold value. The threshold value was defined as the
mean value of Levenshtein distances computed on a set of test word pairs.
Therefore, in case of a partially correct answer, the system both increments the score of the player
and displays the correct expected answer.
Expected answer: la chaise
User Input: la chais
Levenshtein: 0.875
System Feedback: ok! la chaise
Score: score+1
Table 1: An Example of System Feedback.
3. The experiment
We now describe the experiment we performed, in order to study the impact of our 3D interactive
game-like learning environment on young children acquisition of second language vocabulary skills.
3.1 Test Group
The test group was composed of 6 children (9 years old) that voluntarily accepted to participate in the
th
experiment. All the children were in the 4 grade of German primary school and had the same
background knowledge in French (i.e., 1 year school-French). Further, the experimental subjects were
all German native speakers with a poor computer literacy, given the young age.
3.2 Experiment Setting
The children played with a simplified version of the I-Fleg game. Each children played 2 game
sessions. A game session has a duration of 20 minutes and includes a tutorial and a training phase.
During the tutorial phase (5 min.), the children were exposed to both audio output and visual output,
i.e. they heard the pronunciation of the name of the objects they have touched, and further could read
the correct spelling of these words in the tutorial window.
During the training phase (15 min.), when touching an object the children only received audio output
from the system and had to write down the spelling of the name of the objects in the chat box. A
spelling was counted correct if its Levenstein distance to the correct answer was below the set
threshold (in practice, this usually corresponds to cases where the correct answer can be guessed
from the given answer). In case of a wrong answer, the system also displays the expected correct
answer. After playing, the children were asked to name the objects they had seen in the game and if
they had enjoyed playing.
3.3 Evaluation Metrics
For the experiment, we chose a set W of 40 simple French words for the children to learn, 15
providing audio as well as written output and 25 only providing written output. These words all
correspond to objects of the game virtual house that were selected after a pre-test among the words
the children did not yet know. Thus, we divided this set of words into two groups, the set S of objects
that provides written as well as audio output and a second group W of words that only provides written
output.
We then used the following metrics to study the impact of I-Fleg on children learning vocabulary skills
in French and in particular to answer the question whether there is a correlation between the different
types of system output and vocabulary learning.
1)
2)
3)
4)
Ss, the number of words in S the children could right-spell during the game,
Ws, the number of words in W the children could right-spell during the game,
Sr, the number of words in S that the children could remember after the game,
Wr, the number of words in W that the children could remember after the game.
3.4 Results
As the results of the experiments show (see Table 2), the children could remember more easily the
words linked to audio output and this result remains consistent in both game sessions, i.e. the values
of Ss and Sr are always greater than those of Ws and Wr. These results are consistent with the
theories on school-age children learning techniques. It was difficult for these children to focus their
attention on the written output. Children were indeed more interested in exploring the game world,
however they heard the audio output with interest. The results of the question whether they have
enjoyed the game also shows that the game setting was challenging and motivating for the children,
they all want to play again.
Session 1
Player
Game
Session 2
Game after
Game
Game after
Ss
Ws
Sr
Wr
Enjoy
Ss
Ws
Sr
Wr
Enjoy
P1
3
0
3
0
yes
5
0
6
0
yes
P2
1
1
2
1
yes
2
1
2
1
yes
P3
5
0
6
0
yes
7
2
9
4
yes
P4
4
0
6
0
yes
6
1
7
1
yes
P5
6
1
3
1
yes
8
1
9
1
yes
P6
6
2
5
2
yes
7
2
8
2
yes
Table 2. Experimental Results.
Finally a comparison of the results for the two sessions shows that I-Fleg game environment helps
young children to improve vocabulary skills in second language. Most children performed better in the
second game attempt. The mean number of words the children could remember after the second
game session is 8.3. Children in the successive school grades (5th 6th) generally have to learn 20
words in a 45 min lecture. Thus, the achieved results are very encouraging. The different performance
in the two sessions is also due to the fact that children in session 1 had to deal with technical
problems such as learn how to navigate and interact with the virtual game world, find out where the
letters are positioned on the keyboard and further understand the logic of the game.
4. Concluding Remarks
In this paper, we have described an experiment for studying the impact of I-Fleg, an intelligent tutorial
system we have implemented for second language acquisition, on primary school children learning
French as a second language. I-Fleg is designed as a 3D video game-like interactive learning
application. The system version used for the experiment reported in this paper provides written as well
as audio output and is targeted to young learners who want to train French vocabulary. The
experiment is limited in scope and needs to be carried out on a larger scale for the results to be
reliable. Nonetheless, it suggests that two features of I-FLEG are particularly important for language
learning by children. First, audio input is of particular relevance for this target group. We observed that
objects without audio output were largely ignored by children and as the numbers show, they were
also less well learned.
Second, the situated environment provided by the game was also very relevant, stimulating the
children in their learning. The children enjoyed a playful learning experience avoiding to concentrate
on word reading and spelling, which is still a difficult task for primary school children. The preliminary
evaluation that we performed with a small group of children show that the children were motivated by
the playful game setting and that the one-to-one feedback they received from the system enhanced
their learning performance. In future, we plan to evaluate a linguistically more complex version of the
system with adults learners.
References
[1] Aliya, S. K. (2002). The role of computer games in the development of theoretical analysis,
flexibility and reflective thinking in children: A longitudinal study. International Journal of
Psychophysiology, 45, 149.
[2] Marilisa Amoia, Claire Gardent, Laura Perez-Beltrachini. A Serious Game for Second Language
Acquisition. In Proceedings of the 3rd International Conference on Computer Supported Education
(CSEDU 2011). Noordwijkerhout, The Netherlands, 2011.
[3] Boyle, T. (1997). Design for Multimedia Learning, London: Prentice Hall.
[4] M. Csikszentmihalyi, Flow: The Psychology of Optimal Experience, Harper & Row, New York,
1990.
[5] S. Felix 1981. The effect of formal instruction on SLA. Language Learning. 32:87-112.
[6] Johnson, J.S., Newport, E.L. (1989). “Critical Period Effects in Second Language Learning: The
Influence of Maturational State on the Acquisition of English as a Second Language.” Cognitive
Psychology, Vol. 21, pp. 60–99.
[7] Barry McLaughlin, (1985). Second-Language Acquisition in Childhood. Vol. 2: School-Age
Children. Hillsdale, N.J.: Lawrence Erlbaum Associates.
[8] Ministerium für Bildung (Saarland). 2011. Kernlehrplan: Französisch. Grundschule, Klassenstufe
3/4. http://www.saarland.de/dokumente/thema_bildung/KLPFrzGS.pdf
[9] Papert, S. (1993). The Children’s Machine: Rethinking School in the Age of the Computers, New
York: Basic Books.